DNA Methylation Exploration for ARDS: a Multi‑Omics and Multi‑Microarray Interrelated Analysis Shi Zhang, Zongsheng Wu, Jianfeng Xie, Yi Yang, Lei Wang and Haibo Qiu*

Zhang et al. J Transl Med (2019) 17:345 https://doi.org/10.1186/s12967-019-2090-1 Journal of Translational Medicine

RESEARCH Open Access DNA methylation exploration for ARDS: a multi‑omics and multi‑microarray interrelated analysis Shi Zhang, Zongsheng Wu, Jianfeng Xie, Yi Yang, Lei Wang and Haibo Qiu*

Abstract Background: Despite advances in clinical management, there are currently no novel therapeutic targets for acute respiratory distress syndrome (ARDS). DNA methylation, as a reversible process involved in the development and progression of many diseases, would be used as potential therapeutic targets to improve the treatment strategies of ARDS. However, the meaningful DNA methylation sites associated with ARDS still remain largely unknown. We sought to determine the diference in DNA methylation between ARDS patients and healthy participants, and simultaneously, the feasible DNA methylation markers for potential therapeutic targets were also explored. Methods: Microarray data of human blood samples for ARDS and healthy participants up to June 2019 was searched in GEO database. The diference analyses between ARDS and healthy population were performed through limma R package, and furthermore, interrelated analyses of DNA methylation and transcript were accomplished by VennDia- gram R package. Perl and sva R package were used to merge microarray data and decrease heterogeneities among diferent studies. The biological function of screened methylation sites and their regulating genes were annotated according to UniProt database and Pubmed database. GO term and KEGG pathway enrichment analyses were conducted using DAVID 6.8 and KOBAS 3.0. The meaningful DNA methylation markers to distinguish ARDS from healthy controls were explored through ROC (receiver operating characteristic curves) analyses. Results: Five datasets in GEO databases (one DNA methylation dataset, three mRNA datasets, and one mRNA dataset of healthy people) were enrolled in present analyses fnally, and the series were GSE32707, GSE66890, GSE10474, GSE61672, and GSE67530. These databases included 99 patients with ARDS (within 48 h of onset) and 136 healthy participants. Diference analyses indicated 44,439 DNA methylation alterations and 29 diference mRNAs between ARDS and healthy controls. 40 methylation variations regulated transcription of 16 genes was explored via interrelated analysis. According to the functional annotations, 30 DNA methylation sites were related to the imbalance of infammation or immunity, endothelial function, epithelial function and/or coagulation function. cg03341377, cg24310395, cg07830557 and cg08418670, with AUC up to 0.99, might be the meaningful characteristics with the highest perfor- mance to distinguish ARDS from healthy controls. Conclusions: 44,439 DNA methylation alterations and 29 diference mRNAs exist between ARDS and healthy controls. 30 DNA methylation sites may regulate transcription of 10 genes, which take part in pathogenesis of ARDS. These fndings could be intervention targets, with validation experiments to be warranted to assess these further. Keywords: ARDS, DNA methylation, mRNA, Multi-omics, Interrelated analysis

*Correspondence: [email protected] Department of Critical Care Medicine, Zhongda Hospital, School of Medicine, Southeast University, Nanjing 210009, China

© The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Zhang et al. J Transl Med (2019) 17:345 Page 2 of 11

Background we performed integrative analyses of DNA methylation Acute respiratory distress syndrome (ARDS) is a life- variations and mRNA expression diferences. In addition, threatening form of respiratory failure that accounts to evaluate the value of these methylation alterations for 10% of intensive care unit admissions. In spite of to distinguish ARDS, receiver operating characteristic improvements in basic science and clinical research, curves (ROC) would be calculated. treatment options for ARDS are still limited and the mortality of severe ARDS is 40–46% [1]. Most innovative Methods therapies for ARDS have failed in the last decade, and Systematic search and data selection clearly, there is a robust need for better insight in disease Te GEO was searched for all expression microarray that pathogenesis and subsequent emerging treatment strate- matched terms of ARDS. Clinical studies of ARDS using gies [1, 2]. peripheral blood of adult were retained. Datasets that Te majority of studies focus on genomic or transcrip- utilized endotoxin or lipopolysaccharide infusion as vitro tomic responses in ARDS [3–5]. Accumulating evidences or animal models for ARDS were excluded. Clinical-gene demonstrate that the epigenetic alterations especial DNA expression microarray derived from sorted cells was also methylation involved in the development and progres- excluded (Fig. 1). sion of many diseases, including various cancers, lupus, Expression microarray datasets of healthy participants diabetes, asthma, and a variety of neurological disorders were searched and set as the control group that matched [6, 7]. It was reported that DNA methylation can directly terms of health and GPL10558 (the same platform num- block transcription by inhibiting the binding of specifc ber as included gene expression microarray of ARDS). transcription factors to their target sequences on the can- Moreover, where longitudinal data were available for didate gene, which could result in an obvious variation patients admitted with ARDS, we only included data in transcriptome and lead to the occurrence and devel- derived from the within 48 h of onset. opment of diseases eventually [8–10]. In this scenario, valuable DNA methylation variations could be used as biomarkers for molecular diagnosis of disease. Moreo- Gene expression normalization ver, DNA methylation could also be potential target to ComBat normalization in the sva R package and Perl was explore new treatment strategy [6–10]. For instance, used to co-normalize these cohorts into a single cohort, Dhas et al. showed that specifc DNA methylation might after re-normalizing from raw data. All datasets were be potential marker for prognosis of neonatal sepsis downloaded as txt fles, and DNA methylation micro- which may improve the treatment strategies [11]. array were re-normalized and batch corrected through To date, studies on DNA methylation about ARDS minf, impute, and wateRmelon R packages. Outputs is insufcient. Szilagyi et al. showed that myosin light from mRNA array were normal-exponential background chain kinase (MYLK) epigenetic variations were impli- corrected and then between-arrays quantile normalized cated in ARDS pathogenesis and might infuence ARDS using limma R package. For compatibility with microar- mortality [12]. In this study, researchers only focus on ray studies, expression was normalized using a weighted single genetic variations and the other underlying meth- linear regression, and the estimated precision weights of ylated CpGs related genes still unclear. Fortunately, the each observation were then multiplied with the corre- author had uploaded the DNA methylation microar- sponding log2 to yield fnal gene expression values. ray data of patients with ARDS to the Gene Expression Omnibus (GEO) databases, which made it possible for us Integrative analysis to explore DNA methylation in ARDS comprehensively. Tere are two parts to the integrative analysis shown in Beside these, there are three mRNA microarray data of Additional fle 1: Figure S1: on the one hand, the diferen- patients with ARDS in GEO databases which provided by tially expressed DNA methylation sites and mRNAs were Kangelaris et al. [3], Dolinay et al. [4] and Howrylak et al. determined using the R package limma, which imple- [5]. All of these public data will be helpful to us perform ments an empirical Bayesian approach to estimate gene integrative analyses of DNA methylation and mRNA in expression changes using moderated t-tests. A log fold ARDS. change > 0 was defned as hypermethylation or up-regu- In the current study, in order to fnd DNA methylation lated genes, and a log fold change < 0 was defned as dem- alterations and mRNA expression diferences between ethylation or down-regulated genes. For elimination of patients with ARDS and health controls, we utilized dif- the infuence of human species on DNA methylation, the ference analysis on DNA methylation microarray and intersection analyses of the DNA methylation alterations mRNA microarray data. Furthermore, for exploration of of the melanoderm and Caucasian groups were taken, the potentially meaningful DNA methylation alterations, Additional fle 2: Figure S2. Zhang et al. J Transl Med (2019) 17:345 Page 3 of 11

Fig. 1 The fow-process diagram. The fow-process diagram on the left was how to screen microarray datasets

On the other hand, the methylation sites with bio- Te diferentially expressed DNA methylation sites logical function and methylated genes were defned and mRNA were identifed by signifcance criteria as hypermethylation accompanying with down-regu- (adjusted P value < 0.05) as implemented in the R pack- lated genes and demethylation accompanying with up- age limma. Te intersection analyses were performed regulated genes, simultaneously. In this scenario, we through VennDiagram R package. respectively took the intersection of hypermethylation sites and down-regulated genes, and the intersection of demethylation sites and up-regulated genes, Additional fle 3: Figure S3. Zhang et al. J Transl Med (2019) 17:345 Page 4 of 11

Functional annotation and enrichment of screened DNA methylation microarray was the M-values (log2 methylation alterations ratio of the intensities of modifed probe vs unmodi- Functional annotations and gene-annotation enrichment fed probe) of CpG probes (485,577 probes) that passed analyses of genes regulated by screened methylation vari- quality control and batch corrected. Gene microarray ations were referenced the UniProt database, KOBAS 3.0 datasets were mRNA expression profling after quality online database, DAVID 6.8 online database and PubMed control and batch correction, with the median number database. of mRNA probes assayed as 25,128 (ranging from 22,277 For the exploration of the interaction of screened to 47,220). Our integrated dataset included a total of 99 genes, STRING online database was used to plot the pro- patients with ARDS and 136 health participants. Te tein–protein interaction network (PPI). number of samples investigated ranged from 13 to 106 cases (median 30) across the studies. All datasets were Evaluation the value of methylation alterations from United States, and there was no statistical difer- to distinguish ARDS from healthy volunteers ence among datasets in age, and male. To identify novel methylation alterations which would to distinguish ARDS from healthy volunteers, the receiver Screening of diferentially expressed genes operating characteristic curves (ROCs) were performed and methylation sites to calculate the area under the curve (AUC) on screened In total, 22,654 hypermethylation sites and 21,785 dem- DNA methylation variations using ROCR R package. ethylation sites were identifed in Manhattan chart (Fig. 2). Sorted by adjusted P value from small to large, Software and versions the top 10 methylation variations between patients Perl 64 was used to merge data; R x64 3.4.4 was con- with ARDS and healthy volunteers were cg17078393, ducted to process data, analyse data and plot diagrams; cg04794690, cg01564818, cg07748255, cg07369374, Cytoscape 3.6.1 was performed to plot network diagrams. cg23856138, cg21726551, cg25032321, cg10431989 and cg26852712. Obviously, 9 up-regulated mRNA and 20 Results down-regulated mRNA were identifed in volcano plot Characteristics of the datasets (Fig. 2 and Additional fle 4: Table S1). After search strategy and inclusive criteria, 5 studies containing one DNA methylation dataset (GSE67530) [12], Exploration of methylation sites with biological function 3 mRNA datasets of patients with ARDS (GSE32707, and methylated genes GSE66890 and GSE10474) [3–5], and one mRNA data- 16 genes were diferentially methylated by interrelated set of healthy people (GSE61672) [13] were used to build analyses criteria, including 32 hypermethylation sites the methylation and mRNA expression profling datasets, and 8 demethylation sites, demonstrated in Table 2 and Table 1. Fig. 3. Overall, according to the function annotation,

Table 1 Microarray data information Information Transcriptome Transcriptome Transcriptome Transcriptome P Genome ARDS Genome P ARDS 1 ARDS 2 ARDS 3 health health

Sample 18 29 13 106 39 30 Series GSE32707 GSE66890 GSE10474 GSE61672 GSE67530 GSE67530 Platform GPL10558 GPL6244 GPL571 GPL10558 GPL13534 GPL13534 Year 2012 2015 2008 2015 2017 2017 Country USA USA USA USA USA USA Contributor Dolinay T Kangelaris KN Howrylak JA Wingo AP Zhang W Zhang W Age: mean (SD) 54.0 (14.5) 59 (19) 56.2 (4.8) 50.91 (10.4) 0.88 52.5 (17.5) 58.2 (15.1) 0.67 Male% 46.4 55.0 46.0 41.5 0.89 59.0 46.7 0.44 Caucasian% 82.1 NA 84.0 59.4 46.15 50.0 APACHE II score, 30.46 (9.0)a 116 (39)b 20.7(1.5)a 23.5(6.67) mean (SD) Direct lung injury % 50 72 38 100% a APACHE II score b APACHE III score Zhang et al. J Transl Med (2019) 17:345 Page 5 of 11

Fig. 2 Diferential methylation sites and mRNA expression. The Manhattan Figure at the top shows the diferential methylation sites between the patients with ARDS and the healthy controls. The red line shows the 10 sites with minimum P values. The volcano fgure at the bottom shows the mRNA expression. The red points indicate high-expression mRNA in ARDS, and the green points indicate low-expression mRNA according to the threshold of the FDR Zhang et al. J Transl Med (2019) 17:345 Page 6 of 11

Table 2 Relationship between methylation sites and mRNA, with AUC of methylation sites Low expressed Hypermethylation sites AUC to distinguish High expressed Demethylation sites AUC mRNA ARDS mRNA to distinguish ARDS

CX3CR1 cg00262061 0.89 SH3GL1 cg07830557 0.99 CX3CR1 cg03341377 0.99 SH3GL1 cg08418670 0.99 CX3CR1 cg05046858 0.87 SLC3A2 cg02838784 0.79 CX3CR1 cg24310395 0.99 SLC7A1 cg21175585 0.83 CYTIP cg19506253 0.97 SLC7A1 cg26117398 0.77 DUSP6 cg06705834 0.66 TBC1D22B cg08784966 0.93 DUSP6 cg06864046 0.62 TBC1D22B cg12800266 0.95 FYN cg00801571 0.93 TBC1D22B cg22902977 0.74 FYN cg01557421 0.8 FYN cg02115050 0.7 FYN cg02755956 0.9 FYN cg02789394 0.98 FYN cg02816367 0.72 FYN cg04657000 0.94 FYN cg07725064 0.95 FYN cg11412876 0.61 FYN cg12636607 0.92 FYN cg17076443 0.6 FYN cg17587997 0.99 FYN cg20706496 0.93 FYN cg21021905 0.91 FYN cg25189764 0.91 OSBPL8 cg24739189 0.83 PI3 cg24476939 0.93 PILRA cg20784591 0.75 POLB cg07069743 0.49 POLB cg18894794 0.63 RCBTB2 cg00750684 0.54 RNF19B cg10014674 0.55 SRPK2 cg13660622 0.74 SRPK2 cg15857470 0.93 TRIM33 cg18199720 0.66

AUC area under the receiver operating characteristic curves

30 methylation alterations were fnally screened to be level of PI3 in the plasma of patients with ARDS was involved in the pathogenesis of ARDS through their signifcantly decreased. potentially regulated 10 genes (Fig. 4), with the mRNA Te dual specifcity phosphatase 6 (DUSP6), DNA pol- expression of 10 potentially regulated genes demon- ymerase beta (POLB); and solute carrier family 3 mem- strated in Additional fle 5: Figure S4. ber 2 (SLC3A2) genes only had in vitro experimental Peptidase inhibitor 3 (PI3), C-X3-C motif chemokine evidence, and paired immunoglobin like type 2 receptor receptor 1 (CX3CR1) and FYN proto-oncogene, Src alpha (PILRA), SRSF protein kinase 2 (SRPK2), ring fn- family tyrosine kinase (FYN) have been confrmed ger protein 1 (RNF1) and tripartite motif containing 33 to be related to the pathogenesis of infammation or (TRIM33) might be related to the pathogenesis of ARDS, immunity, endothelial function, epithelial function, and as shown in Additional fle 6: Table S2. coagulation function in both ARDS animal models and Gene Ontology (GO) and pathway enrichment analy- in vitro experiments. It was noteworthy that the mRNA ses indicated that the 10 screened genes were enriched Zhang et al. J Transl Med (2019) 17:345 Page 7 of 11

Fig. 3 Circle diagram for the interrelated analysis and PPI analysis results. This diagram shows the positions of methylation sites and their target genes on 24 chromosomes after screening by preliminary interrelated analysis between DNA methylation and mRNA microarray datasets. The red lines represent the relationship between DNA methylation and their target genes. The blue lines represent interactions between proteins coded by these genes according to Protein Interaction Analyses

in 9 functions and 25 pathways, as shown in Additional Assessment of diagnostic efcacy on screened methylation fle 7: Figure S5, Additional fle 8: Figure S6. Tese sites functions and pathways included 2 types. Te frst type AUC was calculated on screened DNA methylation was related to endothelial and epithelial apoptosis and alterations based on ROC analyses, Table 2. Te AUC on repair, including mTOR and MAPK pathways. Te sec- cg03341377, cg24310395, cg07830557 and cg08418670 ond type was related to infammation or immunity, were up to 0.99, which meant that these DNA methyla- including adaptive immune response and leukocyte tion alterations could be most signifcant characteristics migration. to distinguish ARDS from control subjects. Zhang et al. J Transl Med (2019) 17:345 Page 8 of 11

Fig. 4 Network diagram for the relationship between the screened genes and the 4 mechanisms of ARDS. Colour depth represents the log-fold change of genes. Function represents the 4 mechanisms of ARDS. Genes in cluster A was confrmed to be related to the pathogenesis of ARDS in both ARDS animal models and in vitro experiments. Genes in cluster B was confrmed to be related to the pathogenesis of infammation or immunity, endothelial function, epithelial function, and coagulation function in in vitro experiments. Genes in cluster C might be related to the above mechanism because these genes are related to immune regulation, angiogenesis and the cell cycle

Discussion methylation is a reversible process. For instance, in mye- DNA methylation has been most studied and is classi- lodysplastic syndrome, demethylating agents have been cally associated with gene silencing via hypermethylation tested in phase I clinical trials [14]. However, DNA meth- of CpG islands located in promoter regions of various ylation research on ARDS was inadequate and recent diseases suppressor genes. In addition, DNA methylation studies only focused on single genetic variations [15, 16]. is an attractive investigative tool for the study given that Detailed studies of other novel candidates might lead to Zhang et al. J Transl Med (2019) 17:345 Page 9 of 11

identifcation of unsuspected evolutionarily conserved studies, with the function of their methylated gene mechanisms triggered by ARDS. related to immunity and cell proliferation, which could To our knowledge, this was the frst study of all- be appealing to researchers for further studies [25–29]. encompassing analysis for DNA methylation alterations Tejera et al. [30] and Wang et al. [31] had confrmed in ARDS, indicated in Additional fle 9: Table S3. As fur- that the mRNA level of PI3 in patients with ARDS was thermore research targets for improvement of therapies signifcantly decreased, which in turn might increase in ARDS, the present study uncovered 30 methylation the activity of neutrophil elastase indirectly for PI3 alterations and their regulating DNA in ARDS regions. plays a central role in controlling the excessive activity Meanwhile, as diagnostic molecules for ARDS, the clini- of neutrophil elastase. Liu and his colleagues [32] had cians might be interested in the top 10 diference meth- showed that CX3CR1 was the crucial molecule that ylation sites and 4 high diagnostic efcacy CpG islands regulates the EGFR, Src and FAK pathways, which are variations. crucial for epithelial and endothelial cell growth. Te Due to microarray experiments from human blood reduction of CX3CR1 in patients with ARDS could lead samples as the object of this study, 7 animal model stud- to damage to epithelial and endothelial cell regenera- ies and 10 studies not on blood samples were eliminated, tion. However, the mechanism of down-regulated PI3 leaving only 4 studies of microarray experiments and 1 and CX3CR1 in ARDS is unclear. Our study showed study of transcriptome sequencing. In the consideration that the hypermethylation on cg24476939 of PI3 and of comorbidity of hematologic malignancy existing in 4 hypermethylation sites of CX3CR1 might explain patients of transcriptome sequencing study, the sequenc- down-regulation of PI3 and CX3CR1 in ARDS. Similar ing data were eventually removed. Previous studies uti- to PI3 and CX3CR1, FYN, DUSP6, POLB and SLC3A2, lized only one single database, with a relatively small PILRA, SRPK2 and RNF1 were confrmed to be related sample size. Transcriptome microarray datasets were not to the pathogenesis of ARDS, including infammation found healthy participants, which means in the need of or immunity, endothelial–epithelial barrier, and coag- searching healthy volunteers data. Te data of healthy ulation function [33–36]. Te present study verifed participants (generalized anxiety disorder score = 0) from above variations on mRNA levels in ARDS. Moreover, one dataset related to anxiety disorder was extracted, methylation alterations in these DNA regions of mol- according to matched terms of health and GPL10558 ecules might be responsible for the mRNA level vari- (the same platform number as included gene expression ation in ARDS, which might be potential therapeutic microarray of ARDS) in consideration of compatibility targets for ARDS. with microarray studies. In addition, the ROC analyses on screened methyla- After merging these cohorts into a single cohort by tion alterations indicated that the AUC on cg03341377, Perl, batch correction and co-normalization were per- cg24310395, cg07830557 and cg08418670 was up to formed through ComBat normalization in the sva R 0.99, which meant that these DNA methylation altera- package, which was a frequently-used method to decease tions were valuable targets to distinguish ARDS from heterogeneity among microarray studies [17–20]. In healthy controls, Table 2. addition, the threshold values of screening diferential Tere are several limitations to the present study. mRNAs were P value < 0.05 in previous studies, which First, as a retrospective study of primarily publically might lead to partially false positives. Te signifcance available data, we are not able to control for demo- criteria of adjusted P value < 0.05 aimed for reduction of graphics, infection, patient severity, or individual treat- false positives [21, 22]. Strikingly, it has been reported ment, and there was no analysis combined with clinical that the most epigenetic mutations appear unrelated to data because the microarray data did not provide clini- biological function [23]. cal information. Second, although 4 microarray data- Since DNA methylation could directly block tran- sets were merged, a larger sample size may have been scription by inhibiting the binding of specifc transcrip- better. Te ideal method of transcriptome and genome tion factors to their target sequences on the candidate interrelated analysis should measure DNA methylation gene in the result of an obvious variation in transcrip- and mRNA in the same samples and then perform a co- tome, integration with methylation data and matched expression analysis. However, in the public database, transcriptome data was conducted to screen meaning- there was no such completed microarray data regarding ful methylation mutations and methylated DNA regions ARDS. Besides, for the limitation of bioinformatics, the [24]. further studies and experiments should be conducted Until now, the most significant 10 methylation vari- to sequentially verify and research these methylation ations found through difference analyses were lack of alterations screened by our analyses. Zhang et al. J Transl Med (2019) 17:345 Page 10 of 11

Supplementary information Availability of data and materials All data generated and/or analysed during this study can be found in GEO Supplementary information accompanies this paper at https://doi. database. The series of these studies were GSE32707, GSE66890, GSE10474, org/10.1186/s12967-019-2090-1. GSE61672 and GSE67530.

Additional fle 1: Figure S1. The fow-process diagram of the integrative Ethics approval and consent to participate analysis for exploration of the potentially meaningful DNA methylation This study was reviewed and approved by the Institutional Ethics Committee alterations. of Zhongda Hospital. Institutional Ethics Committee of Zhongda Hospital classifed it as exemption from Review. All the participants provided written Additional fle 2: Figure S2. The Venn diagram for the intersection of informed consent. methylation sites between Melanoderm and Caucasian. The left Venn diagram was the intersection of hypermethylation sites between Mel- Consent for publication anoderm and Caucasian. The right Venn diagram was the intersection of Not applicable. demethylation sites between Melanoderm and Caucasian. Additional fle 3: Figure S3. The Venn Diagram for the intersection of Competing interests methylation alterations and mRNA variations. The left Venn Diagram was The authors declare that they have no competing interests. the intersection of the genes between demethylation sites and high expression mRNA. The right Venn Diagram for the intersection of the Received: 9 April 2019 Accepted: 5 October 2019 genes between hypermethylation sites and low expression mRNA. Additional fle 4: Table S1. Log fold change, P value and adjusted P value of diferential mRNA. Additional fle 5: Figure S4. The Violin Diagram for mRNA expression of References 10 screened genes. 1. Fan E, Brodie D, Slutsky AS. Acute respiratory distress syndrome: advances in diagnosis and treatment. JAMA. 2018;319(7):698–710. Additional fle 6: Table S2. Annotations link in UniProt database and 2. Thompson BT, Chambers RC, Liu KD. Acute respiratory distress syndrome. Search strategy in Pubmed database for screened genes. N Engl J Med. 2017;377(19):1904–5. Additional fle 7: Figure S5. The Bar Diagram for GO term enrichment 3. Kangelaris KN, Prakash A, Liu KD, et al. Increased expression of neutrophil- analyses of 10 screened genes. related genes in patients with early sepsis-induced ARDS. Am J Physiol Additional fle 8: Figure S6. The Bubble Diagram for KEGG pathway Lung Cell Mol Physiol. 2015;308(11):L1102–13. enrichment analyses of 10 screened genes. 4. Dolinay T, Kim YS, Howrylak J, et al. Infammasome-regulated cytokines are critical mediators of acute lung injury. Am J Respir Crit Care Med. Additional fle 9: Table S3. Our study provides ARDS researchers with 2012;185(11):1225–34. fve valuable research points. 5. Howrylak JA, Dolinay T, Lucht L, et al. Discovery of the gene signa- ture for acute lung injury in patients with sepsis. Physiol Genomics. 2009;37(2):133–9. Abbreviations 6. Koch A, Joosten SC, Feng Z, et al. Author correction: analysis of ARDS: acute respiratory distress syndrome; GEO: gene expression omnibus; DNA methylation in cancer: location revisited. Nat Rev Clin Oncol. MYLK: myosin light chain kinase; SM: supplemental material; DEGs: diferen- 2018;15(7):467. tially expressed genes; PPI: protein–protein interaction network; GO: gene 7. Vogel CA, LaSalle J. The landscape of DNA methylation amid a perfect ontology; KEGG: Kyoto Encyclopaedia of Genes and Genomes; ROC: receiver storm of autism aetiologies. Nat Rev Neurosci. 2016;17(7):411–23. operating characteristic curves; AUC: area under the curve; PI3: peptidase 8. Bansal A, Pinney SE. DNA methylation and its role in the pathogenesis of inhibitor 3; CX3CR1: C-X3-C motif chemokine receptor 1; FYN: FYN proto- diabetes. Pediatr Diabetes. 2017;18(3):167–77. oncogene, Src family tyrosine kinase; DUSP6: dual specifcity phosphatase 9. Yang IV, Lozupone CA, Schwartz DA. The environment, epigenome, and 6; POLB: DNA polymerase beta; SLC3A2: solute carrier family 3 member 2; asthma. J Allergy Clin Immunol. 2017;140(1):14–23. PILRA: paired immunoglobin like type 2 receptor alpha; SRPK2: SRSF protein 10. Long H, Yin H, Wang L, et al. The critical role of epigenetics in systemic kinase 2; RNF1: ring fnger protein 1; TRIM33: tripartite motif containing 33. lupus erythematosus and autoimmunity. J Autoimmun. 2016;74:118–38. 11. Dhas BB, Antony HA, Bhat V, et al. Global DNA methylation in neonatal Acknowledgements sepsis. Indian J Pediatr. 2015;82(4):340–4. We would like to thank all of the microarray data contributors of this study and 12. Szilagyi KL, Liu C, Zhang X, et al. Epigenetic contribution of the myosin all of the patients and volunteers who participated in this study. light chain kinase gene to the risk for acute respiratory distress syndrome. Transl Res. 2017;180:12–21. Authors’ contributions 13. Wingo AP, Gibson G. Blood gene expression profles suggest altered SZ and HQ had full access to all of the data in the study and take responsibility immune function associated with symptoms of generalized anxiety for the integrity and the accuracy of the data analysis. SZ and HQ performed disorder. Brain Behav Immun. 2015;43:184–91. the data download, bioinformatic analysis, and preparation of the article for 14. Issa JJ, Roboz G, Rizzieri D, et al. Safety and tolerability of guadecitabine publication. Bioinformatics engineer LW checked the methodology, the bio- (SGI-110) in patients with myelodysplastic syndrome and acute myeloid informatic analysis process and the results. All authors participated in writing leukaemia: a multicentre, randomised, dose-escalation phase 1 study. the article and preparing the fgures. All authors read and approved the fnal Lancet Oncol. 2015;16(9):1099–110. manuscript. 15. Huang X, Kong G, Li Y, et al. Decitabine and 5-azacitidine both alleviate LPS induced ARDS through anti-infammatory/antioxidant activity and Funding protection of glycocalyx and inhibition of MAPK pathways in mice. Supported in part by grants from the National Natural Science Foundation Biomed Pharmacother. 2016;84:447–53. of China (Grant Number: 81571847) and the projects of Jiangsu Province’s 16. Lei C, Jiao Y, He B, et al. RIP140 down-regulation alleviates acute lung Medical Key Discipline (ZDXKA2016025). The funding sources had no role in injury via the inhibition of LPS-induced PPARgamma promoter methyla- the design and implementation of the study; the collection, management, tion. Pulm Pharmacol Ther. 2016;37:57–64. analysis, and interpretation of the data; or the preparation, review, or approval 17. Xia L, Su X, Shen J, et al. ANLN functions as a key candidate gene in cervi- of the article. cal cancer as determined by integrated bioinformatic analysis. Cancer Manag Res. 2018;10:663–70. Zhang et al. J Transl Med (2019) 17:345 Page 11 of 11

18. Leek JT, Johnson WE, Parker HS, et al. The sva package for removing batch 30. Tejera P, O’Mahony DS, Owen CA, et al. Functional characterization of efects and other unwanted variation in high-throughput experiments. polymorphisms in the peptidase inhibitor 3 (elafn) gene and validation Bioinformatics. 2012;28(6):882–3. of their contribution to risk of acute respiratory distress syndrome. Am J 19. Leek JT. svaseq: removing batch efects and other unwanted noise from Respir Cell Mol Biol. 2014;51(2):262–72. sequencing data. Nucleic Acids Res. 2014. https://doi.org/10.1093/nar/ 31. Wang Z, Beach D, Su L, et al. A genome-wide expression analysis in blood gku864. identifes pre-elafn as a biomarker in ARDS. Am J Respir Cell Mol Biol. 20. Johnson WE, Li C, Rabinovic A. Adjusting batch efects in microar- 2008;38(6):724–32. ray expression data using empirical Bayes methods. Biostatistics. 32. Liu P, Liang Y, Jiang L, et al. CX3CL1/fractalkine enhances prostate 2007;8(1):118–27. cancer spinal metastasis by activating the Src/FAK pathway. Int J Oncol. 21. Colquhoun D. The reproducibility of research and the misinterpretation 2018;53(4):1544–56. of p-values. R Soc Open Sci. 2017;4(12):171085. 33. Suresh K, Servinsky L, Reyes J, et al. CD36 mediates H2O2-induced calcium 22. Aggarwal S, Yadav AK. False discovery rate estimation in proteomics. infux in lung microvascular endothelial cells. Am J Physiol Lung Cell Mol Methods Mol Biol. 2016;1362:119–28. Physiol. 2017;312(1):L143–53. 23. Shah SP, Roth A, Goya R, et al. The clonal and mutational evolu- 34. Ikeda K, Kinoshita M, Kayama H, et al. Slc3a2 mediates branched-chain tion spectrum of primary triple-negative breast cancers. Nature. amino-acid-dependent maintenance of regulatory T cells. Cell Rep. 2012;486(7403):395–9. 2017;21(7):1824–38. 24. Sanford T, Meng MV, Railkar R, et al. Integrative analysis of the epige- 35. Kobayashi H, Butler JM, O’Donnell R, et al. Angiocrine factors from Akt- netic basis of muscle-invasive urothelial carcinoma. Clin Epigenetics. activated endothelial cells balance self-renewal and diferentiation of 2018;10:19. haematopoietic stem cells. Nat Cell Biol. 2010;12(11):1046–56. 25. Gelkop S, Gish GD, Babichev Y, et al. T cell activation-induced CrkII 36. Vaz M, Hwang SY, Kagiampakis I, et al. Chronic cigarette smoke-induced binding to the Zap70 protein tyrosine kinase is mediated by Lck- epigenomic changes precede sensitization of bronchial epithelial dependent phosphorylation of Zap70 tyrosine 315. J Immunol. cells to single-step transformation by KRAS mutations. Cancer Cell. 2005;175(12):8123–32. 2017;32(3):360–76. 26. Bala S, Kumar A, Soni S, et al. Emp is a component of the nuclear matrix of mammalian cells and undergoes dynamic rearrangements during cell division. Biochem Biophys Res Commun. 2006;342(4):1040–8. Publisher’s Note 27. Jiang M, Gao Y, Yang T, et al. Cyclin Y, a novel membrane-associated cyclin, Springer Nature remains neutral with regard to jurisdictional claims in pub- interacts with PFTK1. FEBS Lett. 2009;583(13):2171–8. lished maps and institutional afliations. 28. Bohnekamp J, Schoneberg T. Cell adhesion receptor GPR133 couples to Gs protein. J Biol Chem. 2011;286(49):41912–6. 29. Kuo CJ, LaMontagne KJ, Garcia-Cardena G, et al. Oligomerization- dependent regulation of motility and morphogenesis by the collagen XVIII NC1/endostatin domain. J Cell Biol. 2001;152(6):1233–46.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions